Reduction of material groups for vertebral bone finite element simulation: cross comparison of grouping methods

In patient-specific biomechanical modeling, the process of image-to-mesh-material mapping is important, and various strategies have been explored for assigning the number of groups of unique material properties to the mesh. This study aims to cross-compare different grouping strategies to identify the minimum number of unique groups necessary for accurately calculating the fracture load of vertebral bones. We analyzed 12 vertebral specimens by experimentally determining the biomechanical fracture load and acquiring corresponding CT scans. After geometry extraction and meshing, we applied commonly used fixed-value strategies for reducing the number of unique groups, such as Modulus Gaping and Percentual Thresholding. Additionally, we introduced a patient-specific adaptive grouping method based on K-means clustering, which allowed us to maintain a consistent number of groups of unique material properties across the study. A total of 204 simulations were performed, achieving a potential 98% reduction in the number of individual material parameters while maintaining a strong correlation with experimental results when utilizing Percentual Thresholding or Adaptive Clustering, compared to Modulus Gaping. The findings demonstrate the feasibility of significantly reducing simulation complexity while maintaining the accuracy of patient-specific models that strongly correlate with experimental results. This reduction enables efficient processing of patient-specific biomechanical models derived from image data, offering potential benefits for clinicians, particularly in resource-constrained settings.